diff --git a/datastore/api/responses.py b/datastore/api/responses.py
index 6f9b748..543e2df 100644
--- a/datastore/api/responses.py
+++ b/datastore/api/responses.py
@@ -1,5 +1,7 @@
 from __future__ import annotations
 
+import base64
+from decimal import Decimal
 from typing import Any
 
 import orjson
@@ -24,6 +26,16 @@
 def _orjson_default(obj: Any) -> Any:
     if hasattr(obj, "model_dump"):
         return obj.model_dump(exclude_none=True)
+    # BigQuery `NUMERIC` / `BIGNUMERIC` columns come back as Decimal —
+    # JSON has no native form, and orjson refuses by default. Stringify
+    # to preserve full precision (NUMERIC = 38 digits, BIGNUMERIC = 76+,
+    # both beyond IEEE-754 double).
+    if isinstance(obj, Decimal):
+        return str(obj)
+    # `BYTES` columns come back as raw `bytes`; base64-encode so the
+    # response stays UTF-8 and round-trippable.
+    if isinstance(obj, bytes):
+        return base64.b64encode(obj).decode("ascii")
     raise TypeError(f"Object of type {type(obj).__name__} is not JSON serializable")
 
 
diff --git a/datastore/services/streaming.py b/datastore/services/streaming.py
index 620fb53..12df494 100644
--- a/datastore/services/streaming.py
+++ b/datastore/services/streaming.py
@@ -29,14 +29,39 @@
 
 from __future__ import annotations
 
+import base64
 import csv
 import io
 from collections.abc import Iterator
+from decimal import Decimal
 from typing import Any
 
 import orjson
 
 
+def _json_default(obj: Any) -> Any:
+    """Serialise types `orjson` refuses out of the box.
+
+    BigQuery `NUMERIC` / `BIGNUMERIC` columns come back as
+    `decimal.Decimal`, which has no native JSON representation.
+    Stringifying preserves full precision (NUMERIC is 38 digits,
+    BIGNUMERIC is 76+ — beyond what a JSON number / IEEE-754 double
+    can represent without loss) and matches CKAN's datastore
+    convention of returning high-precision numerics as strings.
+
+    `bytes` (BigQuery `BYTES` columns) are base64-encoded so the
+    response stays UTF-8 and round-trippable.
+    """
+    if isinstance(obj, Decimal):
+        return str(obj)
+    if isinstance(obj, bytes):
+        return base64.b64encode(obj).decode("ascii")
+    raise TypeError(
+        f"orjson cannot serialise {type(obj).__name__}; "
+        "extend `_json_default` if a new BigQuery type comes through."
+    )
+
+
 def stream_objects(
     *,
     help_url: str,
@@ -219,7 +244,7 @@ def _records_object_array(
             first = False
         else:
             yield b","
-        yield orjson.dumps(dict(zip(columns, row)))
+        yield orjson.dumps(dict(zip(columns, row)), default=_json_default)
     yield b"]"
 
 
@@ -232,7 +257,7 @@ def _records_array_array(records: Iterator[tuple]) -> Iterator[bytes]:
             first = False
         else:
             yield b","
-        yield orjson.dumps(list(row))
+        yield orjson.dumps(list(row), default=_json_default)
     yield b"]"
 
 
diff --git a/tests/test_streaming.py b/tests/test_streaming.py
new file mode 100644
index 0000000..4377f8b
--- /dev/null
+++ b/tests/test_streaming.py
@@ -0,0 +1,81 @@
+"""Regression tests for the streaming row writers in `services.streaming`.
+
+Targets the BigQuery scalar types orjson refuses by default:
+
+  - `NUMERIC` / `BIGNUMERIC` → `decimal.Decimal`
+  - `BYTES`                  → `bytes`
+
+The fix lives in `_json_default` (passed via `orjson.dumps(default=...)`).
+Without it, the stream crashes mid-row with
+`TypeError: Type is not JSON serializable: decimal.Decimal`.
+"""
+
+from __future__ import annotations
+
+import json
+from decimal import Decimal
+
+from datastore.services.streaming import (
+    _records_array_array,
+    _records_object_array,
+)
+
+
+def _join(parts: list[bytes]) -> str:
+    """Stitch the yielded chunks together as a UTF-8 string."""
+    return b"".join(parts).decode("utf-8")
+
+
+def test_records_object_array_serialises_decimal_and_bytes() -> None:
+    """Rows with NUMERIC (Decimal) + BYTES values must stream without
+    blowing up; Decimal is stringified (preserves precision); bytes is
+    base64-encoded."""
+    rows = iter(
+        [
+            ("DCL", Decimal("47.82"), b"\x00\xff"),
+            ("DCH", Decimal("0.00000000000000000000000000000000000001"), b"abc"),
+        ]
+    )
+    columns = ["product_code", "clearing_price_gbp_per_mwh", "signature"]
+
+    body = _join(list(_records_object_array(columns, rows)))
+    records = json.loads(body)
+
+    assert records == [
+        {
+            "product_code": "DCL",
+            "clearing_price_gbp_per_mwh": "47.82",
+            "signature": "AP8=",                 # b64("\x00\xff")
+        },
+        {
+            "product_code": "DCH",
+            "clearing_price_gbp_per_mwh": "1E-38",
+            "signature": "YWJj",                 # b64(b"abc")
+        },
+    ]
+
+
+def test_records_array_array_serialises_decimal_and_bytes() -> None:
+    """Same coverage for `records_format=lists`."""
+    rows = iter([("DCL", Decimal("47.82"), b"\x00\xff")])
+
+    body = _join(list(_records_array_array(rows)))
+    records = json.loads(body)
+
+    assert records == [["DCL", "47.82", "AP8="]]
+
+
+def test_unsupported_type_still_raises() -> None:
+    """We don't want the default to silently swallow new unknown types —
+    bail loudly so the bug surfaces in tests instead of in production."""
+
+    class Mystery:
+        pass
+
+    rows = iter([(Mystery(),)])
+    try:
+        list(_records_array_array(rows))
+    except TypeError as e:
+        assert "Mystery" in str(e)
+    else:
+        raise AssertionError("expected TypeError for unsupported type")